[Pyrex] gcc's -fstrict-aliasing optimization can break vtable construction in PyMODINIT_FUNC

[Dan Gindikin]

Greg Ewing wrote:
> doesn't Python extension code need to be compiled with 
> -fno-strict-aliasing anyway for other reasons?

Not sure, I found this PEP

   http://www.python.org/dev/peps/pep-3123/

didn't find anything in the docs on the matter.

I heard of people doing this hack to get around this problem, have a 
noop function like so:

    void noop(void*) {}

Then if you call noop(vtableptr), that forces the compiler to not 
re-order writes from before and after this call, since it has no way of 
knowing whether "noop()" has any side-effects. Well, unless it inlines 
"noop()".

Dan